Data Cleanup & Audit — Traceable Spreadsheet Reconstruction

Self-directed technical demonstration using synthetic data; not a paid client engagement.

Starting state

180 supplier-price rows with duplicate keys, inconsistent whitespace and case, mixed country, currency and date formats, invalid quantities, and missing lead times.

Transformation

Raw input was preserved. Deterministic rules normalized fields, composite keys identified duplicates, invalid values entered an exception queue, and every removal or change remained traceable.

Delivered state

150 standardized, import-ready rows; 30 duplicate removals with an audit trail; five editable sheets covering raw data, rules, cleaned output, QA actions, and reconciliation.

Quality boundary

Unverifiable values were never guessed. They remained explicitly unresolved, so a human or downstream system can distinguish evidence, transformation, and uncertainty.

The principle is simple: a dataset is not clean merely because it looks uniform. It is clean when material changes can be inspected and reproduced.